AI safety staff take sick leave, quit over fears of technology they built

Researchers at the forefront of artificial intelligence development are taking sick leave, seeking counseling, and resigning in growing numbers as the psychological weight of their work becomes unbearable, according to a Financial Times investigation published on September 22.ft+1

Mental Health Crisis Inside AI Labs

Staff at the UK's AI Safety Institute, the government body tasked with evaluating frontier AI models, are among those hardest hit. Several employees are currently on sick leave and receiving psychological counseling due to stress tied to grueling testing schedules and what one former employee described as a "stressful and, at times, toxic" atmosphere. The strain has fallen most heavily on cybersecurity and biochemistry evaluation teams, whose work involves probing whether AI models can help generate novel biological threats or discover unknown software vulnerabilities.aiweekly+1

The problems extend well beyond the UK institute. Employees at Anthropic, OpenAI, and Google Alphabet Inc. ↘ 3.58% DeepMind are reporting similar difficulties. One DeepMind employee told the Financial Times that AI safety research carries such a high risk of exhaustion it feels like an "impossible mission," adding: "The sense of having lost control, along with the distress of building something that causes real harm, is leading many people to resign or take sick leave".sedaily+1

High-Profile Departures Pile Up

More than a dozen senior researchers have resigned from the three leading Western AI labs over the past two years, according to the Financial Times. The most prominent recent departure was Jacob Coxon, 27, who left Anthropic in early September after just four months, forfeiting unvested equity worth a substantial sum. In a post on X that has drawn over 115 million views, Coxon wrote: "The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt".axios+2

In an interview with CNN, Coxon said AI companies are "gambling with our lives" and warned that the prospect of AI making itself more intelligent is "most scary". His departure was followed by supportive comments from Anthropic alignment researcher Evan Hubinger, who wrote that he personally estimates a greater than 10 percent probability of AI causing human extinction within the next decade.cnbc+1

Calls for Workers to Act

The wave of resignations has drawn attention from academia. At a Stanford HAI event on September 21, Rob Reich, a professor of political science, urged employees at frontier AI companies to consider collective action. Reich noted that during OpenAI's 2023 board crisis, employee solidarity saved Sam Altman's job. "Labor organizing, in this ordinary sense, can also be very consequential across labs to try to pace the frontier," he said, suggesting that if the more than 1,300 signatories of a recent open letter calling for slower AI development haven't considered even a one-day withdrawal of labor, "then I'm not sure I take the whole letter seriously in the first place".

The remarks come amid a broader debate over whether calls for a slowdown from AI executives like Anthropic CEO Dario Amodei and OpenAI's Sam Altman reflect genuine concern or, as tech journalist Steven Vaughan-Nichols argued in Computerworld International Business Machines Corporation ↗ 0.60%, serve to buy time for companies whose finances remain precarious.futurism